Research On Rotating Machinery Fault Diagnosis Method Based On Hierarchical Linked Infinite Hidden Markov Model

This dissertation is supported by the Project supported by the National Natural Science Foundation of China(Grant No.52075236),the Project supported by the Natural Science Foundation of Jiangxi Province,China(Grant No.20212ACB202005),the Project supported by the equipment Pre-Research Foundation of China(Grant No.6142003190210)and the Project supported by the Aviation Science Foundation(Grant No.201946030001).A new fault diagnosis method of rotating machinery based on hierarchically linked infinite Hidden Markov Model(hi HMM)has been deeply studied.The research of this paper has achieved great innovation results.At the same time,the main contents of the paper are as follows:Chapter One discusses the proposition and research significance of this topic.starting from the development of the hidden Markov model and the research status of the hidden Markov model in mechanical fault diagnosis extend to the proposal and application of the hierarchically linked infinite Hidden Markov Model status quo.At the same time,the main content and innovation of this paper are also described.The second chapter discusses the inadequacy of multi-dataset modeling under the relevant system and the problems existing in its own structure.It is extended from the traditional Hidden Markov Model to the infinite Hidden Markov Model,and then to the hierarchically linked infinite Hidden Markov Model.On this foundation,the construction of the hi HMM is clearly stated.In hi HMM,multiple split i HMMs are integrated.By the parameters of i HMM are coupled together to model data sets in related fields,the selftransition probability between adjacent data sets is also clarified.The structure avoids the creation of other redundancy issues between parameters.The end of this chapter also analyzed the inherent advantages of hi HMM in the field of fault diagnosis.In the third chapter,according to the excellent characteristics of the hierarchically linked infinite Hidden Markov Model,it is applied to the rotating machinery rolling bearing and rotor speed-up process.A fault identification method based on the hierarchically linked infinite Hidden Markov Model is proposed.In the proposed method,a hierarchically linked infinite Hidden Markov rolling bearing fault diagnosis model and a hierarchically linked infinite Hidden Markov rotor speed-up fault diagnosis model are established.Input the fault signal features for identification to the model.As the training recognition result of the classifier,the training recognition result obtained by the hidden Markov model as the classifier is compared.The experimental results show that the proposed hi HMM fault diagnosis method is not only obviously superior to the HMM fault diagnosis method and the i HMM fault diagnosis method.The,but also overcomes the shortcomings of the other two methods.In chapter four,the relationship between feature extraction,feature selection and recognition model are deeply analyzed.A fault diagnosis method based on infinite feature selection and improved hierarchically linked infinite Hidden Markov Model is proposed.In the proposed method,after the feature extraction is completed by using the multi-scale arrangement entropy after optimizing the parameters.The obtained feature quantities are sorted by the infinite feature algorithm.Filter out the relatively valuable information and input it into hierarchically linked infinite Hidden Markov Model to train and identify.At the same time,it is compared with the training recognition results which are the infinite Hidden Markov fault diagnosis model trained by using the infinite feature algorithm to filter the features,and the training recognition results which are the infinite Hidden Markov fault diagnosis model trained by using random feature selection.The experimental results show that the improved method selected by the infinite feature algorithm can extract more valuable feature information more effectively.The fifth chapter,the performance degradation trend prediction of key mechanical components is an important research direction in fault diagnosis.The method for predicting the trend of equipment performance degradation is applied to the performance degradation trend of rolling bearings.The proposed method includes four parts:degradation feature extraction,prediction model establishment,degradation state classification and remaining life evaluation.Using the characteristic that the amplitudesensing arrangement entropy is sensitive to the amplitude vibration signal.The whole process of the entire rolling bearing from the normal operation to the degradation and then to the damage is extracted.Input the degradation data of the whole process into the established hierarchically linked infinite Hidden Markov rolling bearing degradation model for training and learning.Use the Viterbi algorithm to reverse the hidden state set of the rolling bearing life data.According to the hidden state set delineate the different states of the bearing.The remaining lifetime and duration are calculated in the set of hidden states.Finally,The experimental results verify the effectiveness of the proposed prediction method.Chapter 6 gives a comprehensive discussion and summary of the research work of this paper.On this basis,it also looks forward to the work content that needs further research.